Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!europa.chnt.gtegsc.com!news.sprintlink.net!cs.utexas.edu!convex!news.duke.edu!godot.cc.duq.edu!newsfeed.pitt.edu!dbisna.com!psinntp!psinntp!psinntp!psinntp!news.rlcn.rl.af.mil!fester.se.rl.af.mil!user
From: marcoj@ai.rl.af.mil (James D. Marco)
Subject: Re: Human brain sized neural networks
Message-ID: <marcoj-0307950507420001@fester.se.rl.af.mil>
Sender: news@news.rlcn.rl.af.mil
Nntp-Posting-Host: fester.se.rl.af.mil
Organization: Rome Laboratory USAF
References: <Pine.ULT.3.91.950628115905.12283C-100000@rac5.wam.umd.edu> <884341210wnr@chmqst.demon.co.uk> <3sus13$6cm@PEAK.ORG> <3t2b11$5jk@news.paonline.com> <3t3qdk$hrm@PEAK.ORG> <3t79hk$4du@news.paonline.com>
Date: Mon, 3 Jul 1995 10:07:42 GMT
Lines: 47

In article <3t79hk$4du@news.paonline.com>, greg.miller@shivasys.com (Greg
Miller) wrote:

>      This is back to what my original question was.  Is the way these
> chemicals perform in the brain effectively represented in neural nets
> today?  If not, how difficult would it be to encode them?  (and has
> anyone ever tried?)

Interesting.  It seems that we ar dealing with two forms of processing
within the human brain.
            1) The base *Physical* structure...including such things as 
               the number of neurons, signal propagation, chemical agents
               (neurotransmitters, "mood" chemicals, various ions- Na and
                others -, RNA and other chemical based "memories", DNA 
                "templates", availability of cellular foods, etc), neural 
                latency and the physical architecture of dendritic 
                connections. 
         
            2) Logical structures such as variable dendritic architecture,
               learned and learning "pathways", biasing, feedback, etc are
               implemented by, and to some degree under the control of, the
               underlying physical brain. 

The believe the physical and logical components interact in a "greyed" 
scaleing  between the two: Instinct vs Learned or Implicit vs Explicit.

To create an artificial intelegence modeled on the Human Brain, the basic
physical architecture needs to be modeled.  Of course emulation of the 
various chemical influences on a Human Brain are needed, both during 
propagation and backpropagation, for example.

Most neural nets I have looked at do not model the physical brain, except
for some implicit assumptions on the function of base "cellular automata". 

I don't know of anyone attempting this.   

Speculation:
The chemical/physical structure of the human brain, base logical structures
and feedback mechanisms, instincts (preprogrammed responses/other
predispositions), some necessary "core" connectedness, and other initial- 
state componentry of the Human Brain, can probably can be modeled by a GA.
The logical structures, learning and memory are best modeled by a NN.  
Associative relations, cognitive processes, problem solving, creative 
reasoning seem to require elements of both.

Comments?               
                                ?;-)        marcoj@ai.rl.af.mil
